Read the original document by opening this link in a new tab.
Table of Contents
1. INTRODUCTION
2. OVERVIEW OF FEDERATED LEARNING
3. TYPES OF FEDERATED LEARNING
4. APPLICATIONS
4.1 Healthcare
4.2 Transportation
4.3 Finance
4.4 Natural Language Processing
5. TRAINING BOTTLENECKS
6. PRIVACY AND SECURITY CONCERNS
6.1 Membership Inference Attacks
6.2 Data Poisoning Attacks
6.3 Model Poisoning Attacks
6.4 Backdoor Attacks
7. RECENT DEVELOPMENTS IN FL
7.1 One-shot federated Learning
7.2 Incentive Mechanisms
7.3 Federated Learning as a Service
Summary
Federated Learning (FL) is a concept where multiple devices collaboratively learn a machine learning model without sharing their private data under the supervision of a central server. This paper discusses the opportunities and challenges in the FL domain. It covers the overview of Federated Learning, different types of FL frameworks, applications in healthcare, transportation, finance, and natural language processing. The paper also addresses training bottlenecks, privacy and security concerns such as Membership Inference Attacks, Data Poisoning Attacks, Model Poisoning Attacks, and Backdoor Attacks. Recent developments in FL include One-shot federated Learning, Incentive Mechanisms, and Federated Learning as a Service.